[Disksim-users] time scale negative value

[Steve]

Hello,Shahrukh Rohinton Tarapore,
    I ran the disksim simulator under the trace whose time granularity is the same to yours,i.e. micro-second. It should be noted by you that the unit of the internal time in disksim is different from that of the sys time in the system level simulator. The unit in the latter is a second. So you can divide your timestamp by 10e6.
    I hope it will work.

All the best.

　　　　　　　　Steve
　　　　　　　　zgy04 at 126.com
　　　　　　　　2006-09-22


======= 2006-09-22 01:21:28 Shahrukh Rohinton Tarapore Written: =======

>Sorry for my late reply Steve, I've had a procmail bug which has been 
>redirecting all my disksim mail to the wrong directory. To answer your 
>questions:
>
>I've tested disksim using a negative time scale value and it never seems 
>to break. As far as it making sense... I don't think it does. My 
>original problem was the traffic in my traces were time stamped so close 
>together (on the order of 1000th of a millisecond) that disksim could 
>not handle it. Making the time stamp negative has been the only way to 
>get disksim to run to completion on these particular fine-grain traces.
>
>I am using disksim in two ways: 1) as a slave to a system level 
>simulator 2) as a validator to the storage simulation in 1. This problem 
>I'm having only occur when validating via traces.
>
>In the past week I've gone very in depth to manually calculate the 
>request latencies disksim provides when the time stamp is negative. 
>There seems to be no problem in running disksim this way, but I agree it 
>doesn't make sense to do so. I'll just say it works and it is something 
>to try if you ever have to run fine-grain time stamped traces in disksim.
>
>-- shahrukh
>
>Steve wrote:
> > Hello,Shahrukh,
> > First,what I wonder is that even if the negative value makes disksim 
> > work well, does it make sense to you?
> > Second, I advise you to make smooth the trace file (out of time 
> > order)by writing a toolkit. It is not hard.
> > Third, I don't know the way you use disksim. Do you use disksim as a 
> > slave module or play the trace with disksim itself? If the latter, I 
> > have no idea. If the former, I have a thread that the syssim.c has a 
> > little bug with respect to calculatiing time.
> > I hope it is useful to you.
> > All the best.
> > 　　　　　　　　Steve
> > 　　　　　　　　zgy04 at 126.com
> > 　　　　　　　　2006-09-17
> > ======= 2006-09-17 09:35:05 Shahrukh Rohinton Tarapore Written: =======
> >  >Can anyone tell me what happens if I set the time scale parameter to a
> >  >negative value? I tried running a trace with disk requests spaced very
> >  >closely together (with respect to time). When I used positive values
> >  >for the time scale I would get errors that requests in the trace file
> >  >were out of time order. Only when I set the time scale parameter
> >  >negative did disksim run through the entire trace without interruption.
> >  >
> >  >I want to be sure that using a negative value doesn't some how skew my
> >  >results. Thanks.
> >  >
> >  >-- shahrukh
> >  >_______________________________________________
> >  >Disksim-users mailing list
> >  >Disksim-users at ece.cmu.edu
> >  >https://sos.ece.cmu.edu/mailman/listinfo/disksim-users
> >  >
> > = = = = = = = = = = = = = = = = = = = =
> > 　　
>_______________________________________________
>Disksim-users mailing list
>Disksim-users at ece.cmu.edu
>https://sos.ece.cmu.edu/mailman/listinfo/disksim-users
>

= = = = = = = = = = = = = = = = = = = =

　　
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.andrew.cmu.edu/pipermail/disksim-users/attachments/20060922/d0e78dc1/attachment.html>